1. Exploratory Data Analysis of The Vancouver Street Trees Dataset¶

This report was prepared by Sarah McDonald on December 12, 2021, as the final project for a Data Visualization class at the University of British Columbia using a subset of the Vancouver Street Trees Data [] provided.

Fig. 1.1 Street trees in Vancouver¶

# Import libraries needed for this analysis
import pandas as pd
import altair as alt
import json

pandas [] is used to handle data, altair [] is a package used for graphing, and json [] is used to create maps.

# Load in the data and view a subset
trees_url = 'https://raw.githubusercontent.com/UBC-MDS/data_viz_wrangled/main/data/Trees_data_sets/small_unique_vancouver.csv'
trees_df = pd.read_csv(trees_url, parse_dates=['date_planted'])
trees_df.head()

	Unnamed: 0	std_street	on_street	species_name	neighbourhood_name	date_planted	diameter	street_side_name	genus_name	assigned	...	plant_area	curb	tree_id	common_name	height_range_id	on_street_block	cultivar_name	root_barrier	latitude	longitude
0	10747	W 20TH AV	W 20TH AV	PLATANOIDES	Riley Park	2000-02-23	28.5	EVEN	ACER	N	...	15	Y	21421	NORWAY MAPLE	4	0	NaN	N	49.252711	-123.106323
1	12573	W 18TH AV	W 18TH AV	CALLERYANA	Arbutus-Ridge	1992-02-04	6.0	ODD	PYRUS	N	...	7	Y	129645	CHANTICLEER PEAR	2	2300	CHANTICLEER	N	49.256350	-123.158709
2	29676	ROSS ST	ROSS ST	NIGRA	Sunset	NaT	12.0	ODD	PINUS	N	...	7	Y	154675	AUSTRIAN PINE	4	7800	NaN	N	49.213486	-123.083254
3	8856	DOMAN ST	DOMAN ST	AMERICANA	Killarney	1999-11-12	11.0	EVEN	FRAXINUS	N	...	7	Y	180803	AUTUMN APPLAUSE ASH	4	6900	AUTUMN APPLAUSE	N	49.220839	-123.036721
4	21098	EAST BOULEVARD	EAST BOULEVARD	HIPPOCASTANUM	Shaughnessy	NaT	15.5	ODD	AESCULUS	Y	...	N	Y	74364	COMMON HORSECHESTNUT	4	5200	NaN	N	49.238514	-123.154958

5 rows × 21 columns

# get more information about our datasset
trees_df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 5000 entries, 0 to 4999
Data columns (total 21 columns):
 #   Column              Non-Null Count  Dtype         
---  ------              --------------  -----         
 Unnamed: 0          5000 non-null   int64         
 std_street          5000 non-null   object        
 on_street           5000 non-null   object        
 species_name        5000 non-null   object        
 neighbourhood_name  5000 non-null   object        
 date_planted        2363 non-null   datetime64[ns]
 diameter            5000 non-null   float64       
 street_side_name    5000 non-null   object        
 genus_name          5000 non-null   object        
 assigned            5000 non-null   object        
civic_number        5000 non-null   int64         
plant_area          4950 non-null   object        
curb                5000 non-null   object        
tree_id             5000 non-null   int64         
common_name         5000 non-null   object        
height_range_id     5000 non-null   int64         
on_street_block     5000 non-null   int64         
cultivar_name       2658 non-null   object        
root_barrier        5000 non-null   object        
latitude            5000 non-null   float64       
longitude           5000 non-null   float64       
dtypes: datetime64[ns](1), float64(3), int64(5), object(12)
memory usage: 820.4+ KB

2. Questions of Interest¶

For this analysis I am interested in how the number and type of trees planted has changed over time. From our initial look at the data, I can see that a lot of values are missing from the ‘date_planted’ column. This could be an error in data recording or it could be that we don’t have records of when older trees were planted. To visualize the gaps in our data, let’s first plot the dates we do have.

# rug plot to visualize date_planted column data
trees_date = alt.Chart(trees_df).mark_tick().encode(
             alt.X("date_planted:T", scale=alt.Scale())
             )

trees_date

It looks like we have continuous data from 1989-2019. If our theory is correct and data without values in the ‘date_planted’ column is from older trees, we could expect these trees to be larger than trees planted more recently. Let’s see if that holds true for our data.

To make the data easier to filter I will use the pandas [] package to add a new column to our data frame. A simple boolean will let us see if the date planted is availabe for that entry.

{
    trees_nan = trees_df.assign(date_record = trees_df.isna().loc[:,'date_planted'])
}

# add a boolean column to our datafrom for data_planted data available
trees_nan = trees_df.assign(date_record = trees_df.isna().loc[:, 'date_planted'])
trees_nan.head()

	Unnamed: 0	std_street	on_street	species_name	neighbourhood_name	date_planted	diameter	street_side_name	genus_name	assigned	...	curb	tree_id	common_name	height_range_id	on_street_block	cultivar_name	root_barrier	latitude	longitude	date_record
0	10747	W 20TH AV	W 20TH AV	PLATANOIDES	Riley Park	2000-02-23	28.5	EVEN	ACER	N	...	Y	21421	NORWAY MAPLE	4	0	NaN	N	49.252711	-123.106323	False
1	12573	W 18TH AV	W 18TH AV	CALLERYANA	Arbutus-Ridge	1992-02-04	6.0	ODD	PYRUS	N	...	Y	129645	CHANTICLEER PEAR	2	2300	CHANTICLEER	N	49.256350	-123.158709	False
2	29676	ROSS ST	ROSS ST	NIGRA	Sunset	NaT	12.0	ODD	PINUS	N	...	Y	154675	AUSTRIAN PINE	4	7800	NaN	N	49.213486	-123.083254	True
3	8856	DOMAN ST	DOMAN ST	AMERICANA	Killarney	1999-11-12	11.0	EVEN	FRAXINUS	N	...	Y	180803	AUTUMN APPLAUSE ASH	4	6900	AUTUMN APPLAUSE	N	49.220839	-123.036721	False
4	21098	EAST BOULEVARD	EAST BOULEVARD	HIPPOCASTANUM	Shaughnessy	NaT	15.5	ODD	AESCULUS	Y	...	Y	74364	COMMON HORSECHESTNUT	4	5200	NaN	N	49.238514	-123.154958	True

5 rows × 22 columns

To account for differences in species we want to break the records down by species. First let’s see how many species we are working with.

species = trees_nan.groupby("species_name")
species.describe()

	Unnamed: 0								diameter		...	latitude		longitude
	count	mean	std	min	25%	50%	75%	max	count	mean	...	75%	max	count	mean	std	min	25%	50%	75%	max
species_name
ABIES	3.0	11484.666667	7736.631718	4347.0	7374.00	10401.0	15053.5	19706.0	3.0	16.000000	...	49.251689	49.265250	3.0	-123.139497	0.082919	-123.191800	-123.187300	-123.182800	-123.113346	-123.043891
ACERIFOLIA X	60.0	14736.833333	7736.247569	1152.0	8729.25	12926.0	21225.0	29978.0	60.0	22.355000	...	49.263235	49.289708	60.0	-123.117517	0.047075	-123.198230	-123.150238	-123.122816	-123.078775	-123.030066
ACUTISSIMA	19.0	16161.631579	8395.660984	2483.0	11159.00	16611.0	23396.5	28798.0	19.0	11.355263	...	49.263155	49.285991	19.0	-123.087162	0.038076	-123.166011	-123.113721	-123.089016	-123.058271	-123.028403
ALNIFOLIA	7.0	19888.285714	6129.725299	11189.0	15721.50	21692.0	24053.0	26788.0	7.0	7.642857	...	49.271948	49.290517	7.0	-123.086372	0.052193	-123.157361	-123.132622	-123.055624	-123.044008	-123.038358
ALPINUM	1.0	7160.000000	NaN	7160.0	7160.00	7160.0	7160.0	7160.0	1.0	8.000000	...	49.261980	49.261980	1.0	-123.176110	NaN	-123.176110	-123.176110	-123.176110	-123.176110	-123.176110
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
WATERERI X	3.0	10674.000000	4363.166396	7523.0	8184.00	8845.0	12249.5	15654.0	3.0	18.833333	...	49.247132	49.258560	3.0	-123.137358	0.067524	-123.209370	-123.168305	-123.127239	-123.101351	-123.075464
X YEDOENSIS	90.0	16544.900000	8492.142408	832.0	9711.75	17409.5	22845.5	29792.0	90.0	7.547222	...	49.256834	49.289456	90.0	-123.117258	0.057906	-123.220360	-123.166355	-123.130054	-123.058314	-123.025868
XX	57.0	16790.192982	9060.538799	397.0	8913.00	18314.0	25920.0	29855.0	57.0	3.504386	...	49.261244	49.289050	57.0	-123.097158	0.050363	-123.209720	-123.137614	-123.088452	-123.060023	-123.023650
YUNNANENSIS	1.0	5188.000000	NaN	5188.0	5188.00	5188.0	5188.0	5188.0	1.0	10.000000	...	49.220989	49.220989	1.0	-123.100972	NaN	-123.100972	-123.100972	-123.100972	-123.100972	-123.100972
ZUMI	65.0	13045.923077	8931.664130	33.0	5494.00	9988.0	21052.0	29456.0	65.0	5.203846	...	49.264133	49.285638	65.0	-123.101302	0.058736	-123.214080	-123.157532	-123.082970	-123.055951	-123.026981

171 rows × 64 columns

3. Top 10¶

171 is a lot of species to visualize all at once. Let’s find our top 10 using the pandas package [] to group entries by their common name, count those entries, and sort from most to least common. Then we can filter so we see only the first 10 entries, the 10 most common trees planted!

{
    trees_common = (trees_nan.groupby("common_name").count().sort_values(by='tree_id', ascending=False
                    ).reset_index().loc[0:9])

}

#find the 10 most common trees in our dataset
trees_common = (trees_nan.groupby("common_name").count().sort_values(by='tree_id', ascending=False
                ).reset_index().loc[0:9])
trees_common = trees_common["common_name"].tolist()
trees_common

['KWANZAN FLOWERING CHERRY',
 'PISSARD PLUM',
 'NORWAY MAPLE',
 'CRIMEAN LINDEN',
 'PYRAMIDAL EUROPEAN HORNBEAM',
 'NIGHT PURPLE LEAF PLUM',
 'KOBUS MAGNOLIA',
 'AKEBONO FLOWERING CHERRY',
 'RED MAPLE',
 'KATSURA TREE']

# filter trees_nan to include only the most common trees
common_records = trees_nan.common_name.isin(trees_common)
trees_nan_small = trees_nan[common_records]

# chart average tree diameter per species (most common)
tree_diam = alt.Chart(trees_nan_small).mark_boxplot().encode(
            alt.X('diameter:Q'),
            alt.Y('common_name:N'),
            ).properties(width=300).facet('date_record')
tree_diam

As we can see from the chart above, trees without a date record do have a higher median diameter than trees with a date record. Trees increase in circumference as they age, a general formula for estimating the age of a trees is the diameter of the tree multiplied by a growth factor specific to the species.[]

\[ age \approx \frac{C}{\pi} \times G \]

Our theory that trees without date records are older seems be correct, we will exclude these values from future plots regarding date. To make analysis easier, I will add a column with just the year planted.

# remove entries with no date_planted
trees_small = trees_df.dropna(subset=['date_planted'])
# create a new column with just year 
trees_small = trees_small.assign(year_planted = trees_small['date_planted'].dt.year)

# number of trees planted over time
trees_time = alt.Chart(trees_small).mark_bar().encode(
             alt.X('year_planted:O'),
             alt.Y('count()'))
trees_time

4. Click to filter¶

Let’s make this chart clickable so we can filter our top 10 tree species by year.

click_year = alt.selection_multi(encodings=['x'], on='click')
click_trees_year = (trees_time.encode(
                   opacity=alt.condition(click_year, alt.value(1), alt.value(0.5)))
                  .properties(height=100, width=500)
                  .add_selection(click_year))

# select 10 most common trees based on year
species_select = (alt.Chart(trees_small).transform_filter(click_year).mark_bar().encode(
                    alt.Y('species_name:N', sort='x'),
                    alt.X('species_count:Q'),
                    ).transform_aggregate(
                    species_count="count()",
                    groupby=["species_name"]
                    ).transform_window(
                    rank='rank(species_count)',
                    sort=[alt.SortField("species_count", order="descending")]
                    ).transform_filter((alt.datum.rank <= 10)).add_selection(click_year))
species_select & click_trees_year

Interesting, there is less overlap in the top 10 species per year than I thought there would be. Now, I would like to look more at the size of trees. I wonder how the method of planting affects a tree’s size. To visualize I will use our top 10 data subset.

# Tree diameter vs height colored by species
tree_height = alt.Chart(trees_nan_small).mark_circle().encode(
              alt.X('diameter:Q'),
              alt.Y('height_range_id:Q'),
              color='species_name:N'
              )
tree_height

# facet our size chart by root barrier
tree_height.facet('root_barrier:N')

# facet tree size by side of street
tree_side = tree_height.properties(width=200).facet('street_side_name')
tree_side

5. Barriers to barriers¶

It looks like the side of the street trees are planted on makes no difference to size however, trees planted with a root barrier do seem to be smaller. Let’s see if the trees with root barriers are younger than those without using our full dataset.

Fig. 5.1 An example of a root barrier. These are used to prevent tree roots from damaging the sidewalk.¶

root_barrier = trees_time.encode(color="root_barrier:N")
root_barrier

It looks like most of the trees with root barriers were planted between 2004 and 2009. Let’s filter our data to include just those years and see if the pattern still holds.

tree_height_filter = alt.Chart(trees_small).transform_filter(
                   alt.FieldRangePredicate(field='year_planted', range=[2004, 2009])
                   ).mark_circle().encode(
                   alt.X('diameter:Q'),
                   alt.Y('height_range_id:Q')
                   ).properties(width=300).facet('root_barrier:N')
tree_height_filter

When we filter just for years that used root barriers the size difference is much less pronounced. Our initial observations about root barriers could have been because a smaller percentage of the data used root barriers.

Now, lets see how the trees are distributed over Vancouver. As part of this course code was provided to create a base map of Vancouver.

# load data to make a map of vancouver (code provided)
url_geojson = 'https://raw.githubusercontent.com/UBC-MDS/exploratory-data-viz/main/data/local-area-boundary.geojson'
data_geojson_remote = alt.Data(url=url_geojson, format=alt.DataFormat(property='features',type='json'))
data_geojson_remote

Data({
  format: DataFormat({
    property: 'features',
    type: 'json'
  }),
  url: 'https://raw.githubusercontent.com/UBC-MDS/exploratory-data-viz/main/data/local-area-boundary.geojson'
})

# base map of Vancouver (code provided)
vancouver_map = alt.Chart(data_geojson_remote).mark_geoshape(
    color = 'white', opacity= 0.5, stroke='black').encode(
).project(type='identity', reflectY=True)

vancouver_map

#Map location of all trees in Vancouver
points = alt.Chart(trees_small).mark_circle(size=20).encode(
         longitude='longitude',
         latitude='latitude',
         ).project(type= 'identity', reflectY=True)

point_map = (vancouver_map + points)
point_map

To see how the distribution changes over time I am going to use the clickable year chart we made earlier.

point_map = point_map.encode(
                opacity=alt.condition(click_year, alt.value(1), alt.value(0.1)),
                color="species_name:N"
                ).add_selection(click_year)
point_map & click_trees_year

6. Conclusion¶

Interesting, over the years the distribution seems to be spread out evenly. I would have guessed that the street tree program would have started in a few neighbourhoods and branched out from there. There also doesn’t seem to be any clusters of particular species in neighbourhoods but it is hard to tell with so many species to consider. For the analysis report I think it will be interesting to explore the distribution of species planted over time and space using both time charts and a map. Linking our top 10 species per year chart will make the species distribution much easier to visualize. I am also very interested in our findings about the size of trees and root barriers so I will include those in our report as well.

My book - Assignment 8

Exploratory Data Analysis of The Vancouver Street Trees Dataset

Contents